Can AI enhance VR training? A systematic review of AI-VR training research.

Michael DeFabiis1, Kevin Askew1, Valerie Sessa1


1 Montclair State University

Introduction

Very little research exists highlighting the presence or absence of training best practices.

RQ 1: Are the current VR and AI training programs utilizing best practices from the science of training?

RQ 2: Are current studies on AI-enhanced VR training programs utilizing a proper control group?

RQ 3: Do current studies on AI-enhance VR training programs have a sufficient number of people to make inferences about their effectiveness?

RQ 4: What are the preliminary findings? Does adding AI to VR training result in a better experience for the trainee, greater learning, or better on-the-job performance than regular VR training?

Methods

Results

Discussion

(#tab:table 6)Table 6: Guidelines for Future AI-VR Training Research and Adherence to Best Practices.
Research Question Guidelines for Future Research
1a. Are learning objectives created and clarified to the reader? Create clear learning objectives by including a behavior and criterion. The condition is often implied for AI-VR training articles, but it can be helpful to include.
1b. Is the training delivered in a way that builds trainees’ belief in their ability to learn and display trained skills (self-efficacy)? Measure self-efficacy directly using a self-efficacy assessment.
1c. Does the training encourage trainees to participate in training to learn rather than to appear capable (promotes a learning orientation)? Include effective feedback mechanisms, encourage error making, and create training aspects that focus on learning rather than performance. Assessing participants inside of a AI-VR training program can inhibit learning outcomes.
1d. Does the training engage trainees and build their interest? Measure engagement directly using an engagement assessment.
1e. Does the training utilize a valid training strategy and design? This involves providing information, giving demonstrations of good and bad behaviors, allowing opportunities to practice, and providing meaningful feedback. AI-VR training programs should strive to incorporate all of these aspects.
1f. Does the training allow trainees to use the same cognitive processes that they will have to in the environment this learning should transfer to? Create immersive training programs that replicate real life tasks and scenarios.
1g. Does the training keep trainees’ attention by allowing trainees to monitor their progress toward goals? AI-VR training programs can allow trainees’ to monitor progress via user interface (UI) elements such as maps and progress bars. Trainings programs can also implement checkpoints and endpoints with visual / haptic feedback (success screens, animations, sounds, etc.)
1h. Does the training encourage trainees to make errors? Provide instant feedback and additional opportunities for practice to encourage errors. Do not assess participant performance inside of the training program.
1i. Does the training provide sufficient structure to trainees when allowing them to make decisions about their learning experience? Build game elements that allow trainees to select training courses or a preferred order of courses. Create scenarios that allow autonomous movement and decision making. Use algorithms that select training scenarios based on participant’s training decisions.
1j. Does the training simulation increase psychological fidelity (e.g. job-relevant, technology used fits the task)? Ensure that the training programs are job-relevant, have good task-technology fit, and are immersive enough to boost user presence.
1k. Does the training use established learning / outcome taxonomies (e.g. affective, cognitive, and/or behavioral indicators)? Always include cognitive and / or behavioral indicators. Include reaction measures beyond engagement and self-efficacy to help provide more information to readers.
Note:
This table highlights guidelines for achieving each of the training best practices suggested by Salas (2012).

References